Hydrodynamic simulators are used to predict the operation of a production well. Their work demands high performance. Alternative simplified methods are appearing constantly, for example, proxy models. Firstly, these models are tuned to historical data and then they make predictions. The source data is quite inaccurate and incomplete often. It is necessary to study how an input parameter error affects quality of forecast. This article shows an algorithm for determining the well interference coefficients by linear regression method. Study was made of the stability of the obtained solution to the noise of the initial data under various conditions. A production well forecast has been obtained, which is then compared with the exact value. The error in which initial data has the greatest impact on the forecast error is analyzed. Possible options for ensuring sustainability are proposed.